Challenges of deep learning diagnosis for COVID-19 from chest imaging

Abstract The COVID-19 pandemic has spread worldwide for over 2 years now. raises a significant threat to global health due its transmissibility and high pathogenicity. current standard detection method COVID-19, namely, reverse transcription–polymerase chain reaction (RT–PCR), is slow inaccurate help fight the pandemic. RT–PCR takes hours days report single test result false-negative rate. As result, an infected person with negative may unknowingly continue virus. Thus, better methods are required improve control of COVID-19. With technology advancements in artificial intelligence machine learning, deep-learning diagnostic studies detect infection using medical chest imaging have emerged. In this paper, we review these by analyzing their approaches highlighting major challenges. These challenges include dataset cleanness, public availability, capability differentiate from unrelated viral pneumonia, difficulty dealing images multiple points view. Finally, discuss various ideas solutions address highlighted reviewed papers.